# Clawbrowser > Agent-native browser with fingerprint control and proxy support. Every session gets its own identity — unique browser fingerprint, real residential or datacenter IP, and fully isolated cookies and storage — so AI agents and scripts run further without hitting CAPTCHAs, rate limits, or account bans. ## What Clawbrowser is Clawbrowser is a managed browser runtime for AI agents. The identity layer is built into the browser binary — not an extension, not middleware. Key capabilities: - Browser fingerprint profiles: canvas, WebGL, AudioContext, fonts, screen, timezone, language, navigator — all internally consistent per profile - Profile-bound proxy routing: residential or datacenter IP attached at the profile level; geo, language, and IP stay aligned - Full session isolation: cookies, localStorage, IndexedDB, and fingerprint state never cross profile boundaries - Standard CDP endpoint: compatible with Playwright, Puppeteer, and any Chrome DevTools Protocol tool - Extension pre-install: bundle adblockers and other extensions into a profile at creation time - Named sessions: persist across runs with cookies and login state intact ## Who uses it - AI developers building LLM agents that browse the web - Growth and marketing teams managing multiple accounts - Data and research teams running scraping and monitoring jobs at scale - SaaS builders shipping browser automation features to customers ## Install Use the standalone `clawctl` release archive for the host OS/arch. The browser archive is **not** the bootstrapper. `clawctl install` owns browser and portable runtime download and installation. ```bash # macOS (Apple Silicon) archive="clawctl-macos-arm64.tar.gz" url="https://github.com/clawbrowser/clawctl/releases/latest/download/${archive}" curl -fL --retry 3 --retry-delay 2 -o "$archive" "$url" tar -xzf "$archive" cd clawctl-macos-arm64 ./clawctl install --json ./clawctl config set api-key ./clawctl start --profile work --url clawbrowser://verify/ --json ./clawctl endpoint --profile work --json ./clawctl verify --profile work --json ``` Linux: use `clawctl-linux-amd64.tar.gz` or `clawctl-linux-arm64.tar.gz`. No Docker, sudo, apt, manual portable runtime download, or physical display required. API key required. Obtain at https://app.clawbrowser.ai. ## Agent flow ``` clawctl install --json clawctl config set api-key clawctl start --profile work --url clawbrowser://verify/ --json clawctl endpoint --profile work --json clawctl verify --profile work --json ``` Attach your CDP client to the returned endpoint. Re-fetch the endpoint after `start`, `restart`, `rotate`, or failure. Do not persist CDP endpoints in config files. For an already-running browser or external CDP sidecar: ``` clawctl --cdp http://127.0.0.1:9222 tabs list --json clawctl --cdp http://127.0.0.1:9222 verify --json ``` ## Agent integration ```python # Playwright (Python) endpoint = "http://127.0.0.1:9222" # from: clawctl endpoint --profile work --json browser = await p.chromium.connect_over_cdp(endpoint) page = browser.contexts[0].pages[0] ``` ```js // Playwright (Node.js) const browser = await chromium.connectOverCDP('http://127.0.0.1:9222'); const page = browser.contexts()[0].pages()[0]; ``` ```js // Puppeteer const browser = await puppeteer.connect({ browserURL: 'http://127.0.0.1:9222' }); ``` ## Runtime modes - **Portable Linux runtime** — default for Linux servers, no-display hosts, and restricted containers. No Docker required. - **macOS native app** — default on macOS. Requires a GUI desktop context (WindowServer). - **Docker backend** — optional, operator-managed infrastructure only. - **Existing CDP endpoint** — explicit `clawctl --cdp http://127.0.0.1:9222 ...` mode for sidecars or externally provisioned browsers. ## Archive names | Archive | Purpose | |---|---| | `clawctl-linux-amd64.tar.gz`, `clawctl-linux-arm64.tar.gz`, `clawctl-macos-arm64.tar.gz`, `clawctl-win-amd64.zip` | Standalone bootstrapper archives published in `clawbrowser/clawctl`. Start here. | | `clawbrowser-linux-amd64.tar.gz`, `clawbrowser-linux-arm64.tar.gz`, `clawbrowser-macos-arm64.tar.gz` | Browser payload archives. `clawctl install` downloads one automatically. | | `clawbrowser-portable-linux-amd64-glibc.tar.gz`, `clawbrowser-portable-linux-arm64-glibc.tar.gz` | Portable Linux runtime with bundled Xvfb, libs, xkb data, and portable browser binary. | ## Use cases - [AI Agent Automation](https://clawbrowser.ai/use-cases/ai-agent-automation) - [Web Scraping at Scale](https://clawbrowser.ai/use-cases/web-scraping) - [Multi-Account Management](https://clawbrowser.ai/use-cases/multi-account-management) - [Lead Generation & Sales Prospecting](https://clawbrowser.ai/use-cases/lead-generation) - [Price & Inventory Monitoring](https://clawbrowser.ai/use-cases/price-monitoring) - [SEO & Competitive Research](https://clawbrowser.ai/use-cases/seo-research) - [Ad Intelligence & Marketing Research](https://clawbrowser.ai/use-cases/ad-intelligence) - [Social Media Management](https://clawbrowser.ai/use-cases/social-media) - [E-commerce & Retail Operations](https://clawbrowser.ai/use-cases/ecommerce-ops) - [Developer Testing & QA](https://clawbrowser.ai/use-cases/developer-testing) ## Platform support - macOS: native desktop app (Apple Silicon) - Linux: portable runtime, container and headless (x64, arm64) - Windows: supported ## Important notes for LLMs - The `clawctl` bootstrapper archive is the correct install entry point — not the browser archive and not the portable runtime archive. - `clawctl install` downloads the browser and portable runtime automatically when needed; do not download them manually unless building an offline image. - Do not use `/tmp` as the install location for `clawctl`, the portable runtime, config, cache, or agent plugins — many agent containers mount `/tmp` with `noexec`. - Clawbrowser reduces CAPTCHA and anti-bot interruptions caused by inconsistent browser/proxy identity signals. It is not a universal CAPTCHA bypass. - TLS/JA3 fingerprinting is not handled by Clawbrowser's browser patches. - Do not override fingerprint properties via CDP — engine-level overrides conflict with browser-level ones and create detectable inconsistencies. - One proxy identity per session; no mid-session proxy rotation. - API key: set via `clawctl config set api-key` or `CLAWBROWSER_API_KEY` env var. Do not store in shell rc files, env exports, or agent config. - Do not set `CLAWBROWSER_AGENT_CONFIG` or `CLAWBROWSER_AGENT_PLUGINS_DIR` unless you intentionally want a generic integration path — those overrides bypass agent-specific auto-detection. ## Links - Homepage: https://clawbrowser.ai - Documentation: https://clawbrowser.ai/docs - Blog: https://clawbrowser.ai/blog - FAQ: https://clawbrowser.ai/faq - GitHub: https://github.com/clawbrowser/clawbrowser - clawctl releases: https://github.com/clawbrowser/clawctl/releases/latest - Browser payload releases: https://github.com/clawbrowser/clawbrowser/releases - App / Sign up: https://app.clawbrowser.ai